There are exceptions in the natural sciences; there are classes of phenomena that embody substantial variance among individuals in the class. Hurricanes and volcanoes are examples of "type-heterogeneous" concepts in the natural sciences. In these examples, the phenomena are grouped together in terms of a set of crude observable characteristics; it is then a question for research to determine whether there are common structures and causal backgrounds that constitute one or more sub-groups of items within the classification. But more typical types of entities in the natural sciences fall into "natural kinds" with common structural and causal properties.
Consider now what is involved in arriving at knowledge about a complex social reality -- the city of Chicago, for example. The social reality of Chicago is constituted by the social behaviors of the individuals who live in Chicago and the institutions that these individuals populate. Consider some of the topics concerning which we might want to gather knowledge:
- What is the health status of Chicagoans?
- What is the climate for race relations in Chicago?
- What is the standard of living in Chicago?
- What is the rate of economic growth in Chicago?
- How do people in Chicago feel about higher education?
- What is the climate for new-business startup in Chicago?
- How well do the institutions of the mayor's office and the city council work?
- How are the Chicago public schools performing?
Notice that almost all these questions invite us to consider the heterogeneity of the population and its organizations. There is an average level of heart disease in the city. But the average is a poor indicator of any particular person's health, because there are socially significant differences across groups with respect to almost all these questions. So a study of health in Chicago requires that we consider some of the ways in which different groups are affected by a variety of circumstances in ways that systematically affect their health status.
These facts suggest that a statement about a social characteristic of a large population needs to be nuanced, indicating the degree of variation of the characteristic across individuals and groups within the population as well as the most significant sub-populations showing the greatest variance from the group mean behavior. Race, ethnicity, gender, geography, age, income, employment, labor-union membership, and education might be variables that define groups with significantly different measures of the variable of interest. And once we have determined that certain social characteristics (race, income, and union membership, let us say) are associated with the outcome of interest (health status, say), then we are stimulated to ask the causal question: what are the social mechanisms at work that produce the associations that are discovered?